Network Estimation Parameter Sensitivity Analysis

Figure 1. Standardized regression coefficients with Age

grid.arrange(a1,a2,a3, nrow=3)

Figure 2. Standardized regression coefficients for cognitive outcomes with neural flexibility.

  • All regression is controlled for age, sex, education

figure by windown size (omega=1, gamma=1)

figdat=rbind(data.frame(fits1[[1]], ws=30),data.frame(fits1[[2]], ws=20),data.frame(fits1[[3]], ws=40)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S')))

ggplot(figdat %>% mutate(ws=factor(ws)), aes(x=network, y=beta)) + 
  geom_line(aes(group=ws, linetype=ws),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

#figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()

figure by gamma (omega=1, window size=30)

figdat=rbind(data.frame(fits1[[4]], ws=0.2),data.frame(fits1[[5]], ws=0.7),data.frame(fits1[[1]], ws=1),
             data.frame(fits1[[6]], ws=1.2),data.frame(fits1[[7]], ws=1.7)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S'))) %>%
  mutate(gamma=factor(ws), ws=30)


ggplot(figdat, aes(x=network, y=beta)) + 
  geom_line(aes(group=gamma, linetype=gamma),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

#figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()

figure by omega (gamma=1, window size=30)

figdat=rbind(data.frame(fits1[[8]], ws=0.2),data.frame(fits1[[9]], ws=0.7),data.frame(fits1[[1]], ws=1),
             data.frame(fits1[[10]], ws=1.2),data.frame(fits1[[11]], ws=1.7)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S')))%>% 
  mutate(omega=factor(ws)) %>%
  mutate(ws=30)


ggplot(figdat , aes(x=network, y=beta)) + 
  geom_line(aes(group=omega, linetype=omega),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

#figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()

Figure 3. Standardized regression coefficients for moderation of cognitive Aging.

figure by windown size (omega=1, gamma=1)

figdat=rbind(data.frame(fits1[[1]], ws=30),data.frame(fits1[[2]], ws=20),data.frame(fits1[[3]], ws=40)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S')))

ggplot(figdat %>% mutate(ws=factor(ws)), aes(x=network, y=beta)) + 
  geom_line(aes(group=ws, linetype=ws),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()
yvar xvar beta CI95.2.5.. CI95.97.5.. pval p.adjust ws condition network sig sig.adjust
6196 MatReas_medianCorRT.log task.SN 0.119 0.009 0.228 0.034 1.000 30 task SN p<0.05 N.S
6236 MatReas_medianCorRT.log task.Visual 0.138 0.024 0.251 0.017 1.000 30 task Visu p<0.05 N.S
667 NPSpeed_attention rs.MRN -0.117 -0.215 -0.020 0.019 0.806 20 rest MRN p<0.05 N.S
61311 NPReasoning task.AUD 0.119 0.016 0.222 0.023 0.806 20 task AUD p<0.05 N.S
62311 NPReasoning task.Visual 0.115 0.015 0.214 0.024 0.806 20 task Visu p<0.05 N.S
61421 NPMemory task.CON -0.127 -0.223 -0.030 0.010 0.806 20 task CON p<0.05 N.S
61141 NPVocab rs.AUD -0.117 -0.223 -0.012 0.030 0.806 20 rest AUD p<0.05 N.S
6631 NPVocab rs.MRN -0.109 -0.210 -0.007 0.036 0.806 20 rest MRN p<0.05 N.S
6381 MatReas_medianCorRT.log rs.DAN 0.171 0.025 0.316 0.022 0.806 20 rest DAN p<0.05 N.S
6461 MatReas_medianCorRT.log rs.DMN_wocerebellum 0.134 0.003 0.264 0.044 0.806 20 rest DMN_ p<0.05 N.S
61412 NPReasoning task.CON 0.103 0.001 0.206 0.049 0.995 40 task CON p<0.05 N.S

figure by gamma (omega=1, window size=30)

figdat=rbind(data.frame(fits1[[4]], ws=0.2),
             data.frame(fits1[[5]], ws=0.7),data.frame(fits1[[1]], ws=1),
             data.frame(fits1[[6]], ws=1.2),data.frame(fits1[[7]], ws=1.7)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S'))) %>%
  mutate(gamma=factor(ws), ws=30)


ggplot(figdat, aes(x=network, y=beta)) + 
  geom_line(aes(group=gamma, linetype=gamma),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()
yvar xvar beta CI95.2.5.. CI95.97.5.. pval p.adjust ws condition network sig sig.adjust gamma
61421 NPMemory task.CON -0.143 -0.242 -0.044 0.005 0.822 30 task CON p<0.05 N.S 0.7
62021 NPMemory task.SMNh -0.112 -0.216 -0.008 0.035 0.967 30 task SMNh p<0.05 N.S 0.7
62121 NPMemory task.SMNm -0.126 -0.225 -0.028 0.012 0.967 30 task SMNm p<0.05 N.S 0.7
61962 MatReas_medianCorRT.log task.SN 0.119 0.009 0.228 0.034 1.000 30 task SN p<0.05 N.S 1
62362 MatReas_medianCorRT.log task.Visual 0.138 0.024 0.251 0.017 1.000 30 task Visu p<0.05 N.S 1
6314 NPSpeed_attention rs.DAN -0.113 -0.224 -0.002 0.045 0.759 30 rest DAN p<0.05 N.S 1.2
679 NPSpeed_attention rs.SN -0.137 -0.243 -0.031 0.011 0.386 30 rest SN p<0.05 N.S 1.2
6149 NPSpeed_attention task.CON -0.095 -0.187 -0.003 0.042 0.759 30 task CON p<0.05 N.S 1.2
61323 NPMemory task.AUD -0.130 -0.237 -0.023 0.018 0.498 30 task AUD p<0.05 N.S 1.2
61423 NPMemory task.CON -0.125 -0.222 -0.029 0.011 0.386 30 task CON p<0.05 N.S 1.2
62023 NPMemory task.SMNh -0.138 -0.238 -0.037 0.007 0.386 30 task SMNh p<0.05 N.S 1.2
62123 NPMemory task.SMNm -0.098 -0.196 0.000 0.050 0.759 30 task SMNm p<0.05 N.S 1.2
62223 NPMemory task.VAN -0.132 -0.231 -0.033 0.009 0.386 30 task VAN p<0.05 N.S 1.2
61143 NPVocab rs.AUD -0.137 -0.243 -0.031 0.011 0.386 30 rest AUD p<0.05 N.S 1.2
61333 NPVocab task.AUD 0.113 0.004 0.221 0.042 0.759 30 task AUD p<0.05 N.S 1.2
6653 MatReas_NumCor rs.MRN -0.144 -0.272 -0.016 0.028 0.674 30 rest MRN p<0.05 N.S 1.2
61410 NPSpeed_attention task.CON -0.094 -0.188 0.000 0.049 0.947 30 task CON p<0.05 N.S 1.7
61124 NPMemory rs.AUD 0.139 0.035 0.244 0.009 0.947 30 rest AUD p<0.05 N.S 1.7
62254 MatReas_NumCor task.VAN -0.116 -0.221 -0.011 0.030 0.947 30 task VAN p<0.05 N.S 1.7

figure by omega (gamma=1, window size=30)

figdat=rbind(data.frame(fits1[[8]], ws=0.2),data.frame(fits1[[9]], ws=0.7),data.frame(fits1[[1]], ws=1),
             data.frame(fits1[[10]], ws=1.2),data.frame(fits1[[11]], ws=1.7)) %>%
  mutate(condition = factor(substr(xvar,1,1), levels=c('r','t'), labels=c('rest','task')),
         network = substr(gsub('task.','',gsub('rs.','',xvar)),1,4),
         sig=factor(ifelse(pval<0.05,'p<0.05','N.S')),
         sig.adjust=factor(ifelse(p.adjust<0.05,'p.adjust<0.05','N.S')))%>% 
  mutate(omega=factor(ws)) %>%
  mutate(ws=30)


ggplot(figdat , aes(x=network, y=beta)) + 
  geom_line(aes(group=omega, linetype=omega),size=1)+
  geom_errorbar(aes(ymin=CI95.2.5.., ymax=CI95.97.5..,colour=sig), alpha=0.2) + 
  facet_grid(yvar~condition, scales="free_y") + 
  geom_hline(yintercept=0, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()

figdat %>% filter(pval<0.05) %>% kable(., digits=3) %>% kable_classic()
yvar xvar beta CI95.2.5.. CI95.97.5.. pval p.adjust ws condition network sig sig.adjust omega
62231 NPVocab task.VAN 0.109 0.006 0.212 0.039 0.960 30 task VAN p<0.05 N.S 0.7
61962 MatReas_medianCorRT.log task.SN 0.119 0.009 0.228 0.034 1.000 30 task SN p<0.05 N.S 1
62362 MatReas_medianCorRT.log task.Visual 0.138 0.024 0.251 0.017 1.000 30 task Visu p<0.05 N.S 1
61423 NPMemory task.CON -0.138 -0.235 -0.041 0.005 0.891 30 task CON p<0.05 N.S 1.2
62223 NPMemory task.VAN -0.113 -0.212 -0.014 0.026 0.981 30 task VAN p<0.05 N.S 1.2
61763 MatReas_medianCorRT.log task.FPN 0.110 0.001 0.220 0.049 0.981 30 task FPN p<0.05 N.S 1.2
62224 NPMemory task.VAN -0.128 -0.223 -0.032 0.009 0.749 30 task VAN p<0.05 N.S 1.7
61144 NPVocab rs.AUD -0.124 -0.228 -0.020 0.020 0.982 30 rest AUD p<0.05 N.S 1.7
6754 MatReas_NumCor rs.SN -0.189 -0.321 -0.056 0.005 0.749 30 rest SN p<0.05 N.S 1.7
61864 MatReas_medianCorRT.log task.MRN 0.122 0.013 0.230 0.028 0.982 30 task MRN p<0.05 N.S 1.7

Figure 4. Paired t-test between task and resting NF

  • n=187 participants

Mean

grid.arrange(a1,a2,a3,ncol=1)

Variance

### figure by windown size (omega=1, gamma=1)

figdat=rbind(data.frame(fits.var[[1]], ws=30),data.frame(fits.var[[2]], ws=20),data.frame(fits.var[[3]], ws=40)) %>%
  mutate(networks = substr(networks,1,4))

a1<-ggplot(figdat %>% mutate(ws=factor(ws)), aes(x=networks, y=var.ratio)) + 
  geom_line(aes(group=ws, linetype=ws),size=1)+
  geom_errorbar(aes(ymin=ci.1, ymax=ci.2,colour=sig), alpha=0.2)+
  geom_hline(yintercept=1, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()+ 
  theme(legend.position='bottom') + 
  ggtitle('By window size  (omega=1, gamma=1)')

#figdat %>% filter(pvalue<0.05) %>% 
#  select(networks, Cohens_d, diff, text) %>% kable(., digits=3) %>% kable_classic()

### figure by gamma (omega=1, window size=30)
figdat=rbind(#data.frame(fits.mean[[4]], ws=0.2),
             data.frame(fits.var[[5]], ws=0.7),data.frame(fits.var[[1]], ws=1),
             data.frame(fits.var[[6]], ws=1.2),data.frame(fits.var[[7]], ws=1.7)) %>%
  mutate(networks = substr(networks,1,4)) %>%
  mutate(gamma=factor(ws), ws=30)


a2<-ggplot(figdat, aes(x=networks, y=var.ratio)) + 
  geom_line(aes(group=gamma, linetype=gamma),size=1)+
  geom_errorbar(aes(ymin=ci.1, ymax=ci.2,colour=sig), alpha=0.2)+
  geom_hline(yintercept=1, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()+ 
  theme(legend.position='bottom') + 
  ggtitle('By gamma (2indow size=30, omega=1)')


#figdat %>% filter(pvalue<0.05) %>% 
#  select(networks, Cohens_d, diff, text) %>% kable(., digits=3) %>% kable_classic()

### figure by omega (gamma=1, window size=30)
figdat=rbind(data.frame(fits.var[[8]], ws=0.2),data.frame(fits.var[[9]], ws=0.7),data.frame(fits.var[[1]], ws=1),
             data.frame(fits.var[[10]], ws=1.2),data.frame(fits.var[[11]], ws=1.7)) %>%
  mutate(networks = substr(networks,1,4)) %>%
  mutate(omega=factor(ws), ws=30)

a3<-ggplot(figdat, aes(x=networks, y=var.ratio)) + 
  geom_line(aes(group=omega, linetype=omega),size=1)+
  geom_errorbar(aes(ymin=ci.1, ymax=ci.2,colour=sig), alpha=0.2)+
  geom_hline(yintercept=1, colour='red')+
  geom_point(aes(colour=sig, shape=sig), size=3) + 
  theme_base()+ 
  theme(legend.position='bottom') + 
  ggtitle('By omega (2indow size=30, gamma=1)')


#figdat %>% filter(pvalue<0.05) %>% 
#  select(networks, Cohens_d, diff, text) %>% kable(., digits=3) %>% kable_classic()
grid.arrange(a1,a2,a3,ncol=1)